Question 4 sticks. runs fine for hours. then bsods.

Feb 21, 2019
5
0
10
0
running:
i7 8700 @ 3.20ghz
Gigabyte Z370 HD3-OP rev. 1

Had 2 sticks of Corsair Vengeance LPX 4GB DDR4 @ 3200MHz running in slots 2 and 4 (provided you view it as 1,2,3,4) on XMP profile 1 for months with no problems.

Just bought another 2 sticks, same RAM, same speed and inserted them into slots 1 and 3. Everything was fine but about an hour into using the computer I would get a blue screen, with the stopcode MEMORY_MANAGEMENT and then later with another one that I only wrote down as FLTMGR.SYS. Googled around, ran MemTest 64, ran windows memory diagnostic and all came up blank. Some other things I tried were making sure the harddrive had 15% free space on it and reseating the modules. Updating the BIOs was another common solution. I Realised I was on version F1 so upgraded to F11b. I didnt realise this had set XMP back to disabled and my RAM was running at 2133MHz for about 2 hours before crashing again.

Some more research and I noticed that although, in the BIOs, Profile DDR Voltage was set as 1.35V (the same as on the RAM stick), the DRAM voltage in a separate menus showed Auto, followed by 1.200V. I manually set this to 1.35 and again, the blue screen. Also checked each channels sub timings and compared them to the stick, both set to 16 18 18 36.

The fact the computer runs flawlessly, or appears to, for hours at a time then shuts down.. It feels like the solution is so small and just out of reach ;/

also: from what I've read having 4 sticks instead of 2 (even if total amount of ram is the same) increases the load on the memory controller, so I need to increase voltages to that regardless of whether or not my CPU is overclocked? VCCIO/VCCSA or something? Perhaps this is the next thing to try

other possibly relevant info/thoughts: Memory Multiplier Tweaker is set to 1. I've read 2 could be beneficial?-something about memory training.
both interleaving options are auto. could manually enable them?
I believe Command rate is set to 2. I should manually set it to 1?
I also noticed onboard graphics is "Enabled" when I use my own GPU. Supposedly this will generate heat and power for no reason?
DRAM Training voltage is set at 1.20. set this to 1.35 also?

**A lot of these potential fixes might only be necessary if the computer wouldnt POST at all so I don't want to go through things at the risk of making more of a balls up that are obviously not the case (but my knowledge is much to small to know which should be tried and which shouldnt) because you would know much sooner than hours after running the pc.
 
Last edited:

InvalidError

Titan
Moderator
If you ran 3200MT/s RAM at 2133MT/s and still got a crash, that smells like bad RAM to me. How long did you run memtest86 for? I'd recommend 10 complete passes, five at the bare minimum.

Another thing you can try is running your PC with only the new memory installed in the same slots as your old memory to keep the number of variabled down and see if you get crashes that way. If this also crashes, then you know with reasonable certainty that at least one of the new DIMMs is either defective or at least doesn't like your MoBo/CPU.
 
Feb 21, 2019
5
0
10
0
How long did you run memtest86 for?
I didn't run memtest 86, ran memtest64 so I could do it without putting it on a usb etc.

that smells like bad RAM to me
Is that how bad RAM would manifest? running fine then suddenly crashing? I've never had a RAM problem before so not sure

Another thing you can try is running your PC with only the new memory installed in the same slots as your old memory to keep the number of variabled down and see if you get crashes that way.
this is a good suggestion. will try.
edit: currently trying this. will update
 
Last edited:

InvalidError

Titan
Moderator
Is that how bad RAM would manifest? running fine then suddenly crashing? I've never had a RAM problem before so not sure
Depends on the nature and location of the defect. If the defect is intermittent and at a memory address that happens to land on a non-critical memory page most of the time, then it can be days or even weeks before the system crashes. If there is a hard cell fault at a memory address that happens to land in the BIOS's shadow-copy memory area that the PC boots from after the memory controller is initialized and BIOS gets copied to RAM, then you may get no-POST.

When I upgraded my Core2 from 4GB to 8GB, one of the new DIMMs had a single intermittent cell. My computer didn't crash until a few days later, rebooted, crashed again a few more days later and this time, Windows wouldn't boot due to corrupted system files. Ran diagnostic on the HDD thinking the HDD was dying, turned out to be fine. Tested the RAM and 4-5 memtest86 passes later, I got my first single-bit fail. That single intermittent memory cell must have corrupted the file system during defrag. Since then, I always run a minimum of five passes of memtest86 whenever I get new RAM before using my PC for anything else.
 
Feb 21, 2019
5
0
10
0
Alright, ran the new ram in the slots I know aren't faulty. received a new error : FAULTY_HARDWARE-CORRUPTED_PAGE

Now trying one of the new sticks and one of my old (working) sticks..
 
Last edited:

InvalidError

Titan
Moderator
Alright, ran the new ram in the slots I know aren't faulty. received a new error : FAULTY_HARDWARE-CORRUPTED_PAGE
Bad memory can cause just about every error imaginable depending on where the corrupted bits landed. A crash of ANY description where you had none with your original memory is a strong indication that there is a problem with the new memory.
 

jonaharkiay

Prominent
Aug 17, 2017
22
0
510
0
I agree. I manually OC'd my ram. When stress testing, an error occured in the 22nd hour - one single error. Computer didn't BSOD or anything, but even small errors like this can corrupt entire OS's. You should be thankful your ram failed spectacularly and not silently corrupting your OS. Whenever you get new ram or run an XMP profile, run custom Prime95 with min FFT on 512 (or something around there). (Max left default). Set Memory allocated to your total memory minus a little (so like 13GB if u have 16). Set test time to 15 minutes. Then, run for 24 hours MINIMUM. Then run HCI memtest (not memtest86) to 1000%. Finally, run hyperpi or superpi 32M (should be 30 min test). If all pass with 0 errors, you should be good.

Also a little word of warning: Ram behaves differently when there are 2 sticks vs 3 or 4. XMP may be unstable for more than came in the original package (2 sticks). 4 Sticks may not be able to handle that xmp. That is why some people buy ram all at once, and why i will always test my ram whenever i do anything to it.
EDIT: Added a few more memory tests
 
Last edited:
Feb 21, 2019
5
0
10
0
Alright guys. lot of information there for me to digest. Thanks a lot for all your help! Currently the PC has been running for about 15 hours.. A good result considering 2-3 was the most in the previous days. Gonna look into some of the software you both mentioned. It's looking like one troublesome stick.. Will probably have to go for 2x 8gb sticks in future.

You should be thankful your ram failed spectacularly and not silently corrupting your OS
every cloud eh :)

Thanks again. Much appreciated.
 

Similar threads


ASK THE COMMUNITY

TRENDING THREADS